Modifying the CROPGRO-soybean model to improve predictions for the upper Midwest
2004
Pedersen, P. | Boote, K.J. | Jones, J.W. | Lauer, J.G.
The CROPGRO-Soybean model has not been extensively evaluated in the upper Midwest. The objective of this project was to determine if modifications of the CROPGRO-Soybean model would improve predictions in the upper Midwest using three cultivars in five management systems and two planting dates from 1997 to 2000. Version 3.5 of the model was compared with 1998 data and found to underestimate total biomass and grain yield at harvest. Changes in temperature function on leaf expansion rate and base temperature for pod addition improved model performance and decreased root mean square error (RMSE) for biomass at harvest and grain yield from 734 to 707 kg ha-1 and from 410 to 362 kg ha-1, respectively. The modified model was then tested with independent data from 1997, 1999, and 2000. Overall, the model parameters calibrated from 1998 data improved the fit slightly but with higher RMSE values for the three independent years than the 1998 data set. Averaged across the 3 yr, the modified model underpredicted biomass at harvest and grain yield by 14 and 6%, respectively, with RMSE for biomass at harvest and grain yield averaging 1181 and 814 kg ha-1, respectively. The inaccuracy was related to underprediction of early vegetative growth because of the effect of site-specific and planting date-specific differences in temperature on biomass accumulation and leaf area index. It was concluded that the modified parameters improved the accuracy of the CROPGRO-Soybean model for the calibration year but did not significantly improve prediction for the three independent years.
显示更多 [+] 显示较少 [-]